📢📢 "Proteina: Scaling Flow-based Protein Structure Generative Models" #ICLR2025 (Oral Presentation) 🔥 Project page: research.nvidia.com/labs/genair/pr… 📜 Paper: arxiv.org/abs/2503.00710 🛠️ Code and weights: github.com/NVIDIA-Digital… 🧵Details in thread... (1/n)

Introducing All-atom Diffusion Transformers — towards Foundation Models for generative chemistry, from my internship with the FAIR Chemistry team FAIR Chemistry AI at Meta There are a couple ML ideas which I think are new and exciting in here 👇

Our review on machine learning methods to study sequence–ensemble–function relationships in disordered proteins

Flow matching has gained popularity recently which is better, diffusion or flow matching? They are formally equivalent Our purpose is to help practitioners understand and use these frameworks interchangeably -- **regardless of what it’s called**

As we age, we accumulate somatic mutations in our cells (mutation clock) and DNA methylation changes (epigenetic clock). They correlate and interact nature.com/articles/s4358… Nature Aging

Extremely excited to give a glimpse of what we've been working on over the past 2 years: unifying structure-prediction and de-novo generation. Neo-1 is an all-atom, fully generative model

Big news from VantAI: Neo-1 integrates SOTA structure prediction and all-atom molecular generation into a single unified model -- an industry first to the best of my knowledge: vant.ai/neo-1

We’ve been excited about discrete flow matching as an alternative to discrete diffusion for biologics design—but how do you scale it to larger vocabularies and control generation? 🌊🌡️Introducing Gumbel-Softmax Flow Matching from our brilliant Sophia Tang! 👇 📜:

Our paper on prediction of phase-separation propensities of disordered proteins from sequence is now published

1/4 🚀 New #ICLR2025 SPOTLIGHT ALERT Gerstein Lab presents “Enhancing the Scalability & Applicability of Kohn-Sham Hamiltonians”—led by  Yunyang & Z Xia & L Huang & J Zhang & Mark Gerstein. Joint work with Microsoft Research.

End to end differentiable protein structure alignment! Jeanne Trinquier Sergey Ovchinnikov

I am pleased to announce a new version of my RL tutorial. Major update to the LLM chapter (eg DPO, GRPO, thinking), minor updates to the MARL and MBRL chapters and various sections (eg offline RL, DPG, etc). Enjoy! arxiv.org/abs/2412.05265

Now that Google has released a text diffusion model, it's time to read this paper.

ChatGPT now can analyze, manipulate, and visualize molecules and chemical information via the RDKit library. Useful for scientific work across health, biology, and chemistry.

A good read at weekend: a recent benchmark of perturbation analysis with machine learning (arxiv.org/pdf/2408.10609). Perturbation analysis is considered one of the key testbed for virtual cells with genomic foundation models. Although perturbation prediction remains an open

This guy literally made Cursor 10x more useful with this one system

Posting my talk tomorrow at the AI4Chemistry Summit at NYU lectures.gersteinlab.org/summary/AI-for… Lots of new slides on computing small molecule Hamiltonians (from Yunyang) & predicting macromolecular misfolding

For a short ML-focused introduction to this paper, I wrote a short blog post ae-foster.github.io/posts/2025/06/… (3/n)

Congratulations to Yale CS faculty/researchers on receiving the Best Paper Award at COLT 2025! 🎉👏👏 #YaleCS #COLT2025 #AI #MachineLearning

🚀 Launch day! The NeurIPS 2025 PokéAgent Challenge is live. Two tracks: ① Showdown Battling – imperfect-info, turn-based strategy ② Pokemon Emerald Speedrunning – long horizon RPG planning 5 M labeled replays • starter kit • baselines. Bring your LLM, RL, or hybrid